Potential Human Health Effects Associated with Stormwater Discharges

Transcription

1 Excerpted from: Stormwater Quality Management, draft, in preparation Robert Pitt Department of Civil and Environmental Engineering University of Alabama Tuscaloosa, Alabama, USA Potential Human Health Effects Associated with Stormwater Discharges Contents Contents...1 Introduction...2 Human Health Effects of Stormwater...2 Evidence of Sewage Contamination of Urban Streams...3 Fort Worth, TX...4 Inner Grays Harbor, WA...4 Sacramento, CA...4 Bellevue, WA...4 Boston, MA...5 Minneapolis/St. Paul, MN...5 Toronto, Ontario...5 Ottawa, Ontario...6 Birmingham, AL...6 Summary of Inappropriate Sanitary Sewage Discharges into Urban Streams...6 Epidemiological Studies and Human Exposures to Stormwater (after Craun, et al. 1996)...7 Water Contact Recreation and Urban Stormwater...9 Development of Bathing Beach Bacteriological Criteria and Associated Epidemiological Studies EPA Guidance for Recreational Waters, Water Supplies, and Fish Consumption...18 Exposure to Pathogens in Stormwater...21 The Presence and Effects of Salmonella in Urban Runoff...22 The Presence and Effects of Staphylococci in Urban Runoff...23 The Presence and Effects of Shigella in Urban Runoff...23 The Presence and Effects of Streptococcus in Urban Runoff...23 The Presence and Effects of Pseudomonas Aeruginosa in Urban Runoff...23 The Presence and Effects of other Pathogens in Urban Runoff...23 Protozoa in Urban Watersheds...24 Santa Monica Bay Project...25 Proposed New California Recreational Area Bacteria Standards...27 Drinking Water Risks and Urban Stormwater...28 Other Human Health Risks Associated with Protozoa and other Microorganisms...32 Microorganisms Observed In Urban Runoff...33 The Contamination of Groundwater by Stormwater Pathogenic Microorganisms...42 Sources of Bacteria and Pathogens found in Urban Runoff...45 Bacteria Survival in Stormwater...47 Survival of Bacteria in Soil...48 Fecal Coliform to Fecal Strep. Bacteria Ratios...49 Water Body Sediment Bacteria...50 Soil Bacteria Sources...51 Wildlife Sources of Bacteria...51 Effects of Birds on Water Bacteria Concentrations...51 Other Wildlife Bacteria Contributions...52 Feces Discharges from Wildlife...53 Case Study: Investigation of Urban Runoff Pathogen Sources in Ottawa, Ontario...62 Estimated Unit Area Bacteria Yields

2 Mammal and Bird Populations and Bacteria Discharges in the Ottawa Urban Area...64 Summary of Ottawa Case Study...67 References...69 Introduction The main purpose of treating stormwater is to reduce its adverse impacts on receiving water beneficial uses. Therefore, it is important in any urban stormwater runoff study to assess the detrimental effects that runoff is actually having on a receiving water. Urban receiving waters may have many beneficial use goals, including: stormwater conveyance (flood prevention) biological uses (warm water fishery, biological integrity, etc.) non-contact recreation (linear parks, aesthetics, boating, etc.) contact recreation (swimming) water supply With full development in an urban watershed and with no stormwater controls, it is unlikely that any of these uses can be obtained. With less development and with the application of stormwater controls, some uses may be possible. It is important that unreasonable expectations not be placed on urban waters, as the cost to obtain these uses may be prohibitive. With full-scale development and lack of adequate stormwater controls, severely degraded streams will be common. However, stormwater conveyance and aesthetics should be the basic beneficial use goals for all urban waters. Biological integrity should also be a goal, but with the realization that the natural stream ecosystem will be severely modified with urbanization. Certain basic controls, installed at the time of development, plus protection of stream habitat, may enable partial use of some of these basic goals in urbanized watersheds. Careful planning and optimal utilization of stormwater controls are necessary to obtain these basic goals in most watersheds. Water contact recreation, consumptive fisheries, and water supplies are not appropriate goals for most urbanized watersheds. However, these higher uses may be possible in urban areas where the receiving waters are large and drain mostly undeveloped areas. Water Environment & Technology (1996a) reported that the latest National Water Quality Inventory released by the EPA only showed a slight improvement in the attainment of beneficial uses in the nations waters. Urban runoff was cited as the leading source of problems in estuaries, with nutrients and bacteria as the leading problems. Problems in rivers and lakes were mostly caused by agricultural runoff, with urban runoff the third ranked source for lakes, and the fourth ranked source for rivers. Bacteria, siltation, and nutrients were the leading problems in the nations rivers and lakes. Borchardt and Sperling (1996) stressed that many conditions may affect receiving waters from stormwater, specifically physical factors (such as shear stress) and chemical factors (such as oxygen depletion and/or non-ionized ammonia). Pathogens in stormwater are also a significant concern potentially affecting human health. The use of indicator bacteria is controversial for stormwater, as well as the assumed time of typical exposure of swimmers to contaminated receiving waters. However, recent epidemiological studies has shown significant health effects associated with stormwater contaminated marine swimming areas. Protozoan pathogens, especially associated with likely sewage-contaminated stormwater, is also of public health concern. The following discussion summarizes these epidemiological concerns, and presents information concerning the pathogenic characteristics of urban stormwater flows. Human Health Effects of Stormwater There are several mechanisms where stormwater exposure can cause potential human health problems. These include exposure to stormwater contaminants at swimming areas affected by stormwater discharges, drinking water supplies contaminated by stormwater discharges, and the consumption of fish and shellfish that have been contaminated by stormwater pollutants. Understanding the risks associated with these exposure mechanisms is difficult and not very clear. Receiving waters where human uses are evident are usually very large and the receiving waters are affected by many sanitary sewage and industrial point discharges, along with upstream agricultural nonpoint discharges, in addition to the local stormwater discharges. In receiving waters only having stormwater 2

3 discharges, it is well known that inappropriate sanitary and other wastewaters are also discharging through the storm drainage system. These interferences make it especially difficult to identify specific cause and effect relationships associated with stormwater discharges alone, in contrast to the many receiving water studies that have investigated ecological problems that can more easily study streams affected by stormwater alone. Therefore, much of the human risk assessment associated with stormwater exposure must use theoretical evaluations relying on stormwater characteristics and laboratory studies in lieu of actual population studies. However, some site investigations, especially related to swimming beach problems associated with nearby stormwater discharges, have been conducted. This section presents a summary of the human health effects of stormwater discharges, stressing these swimming beach studies. There are several impediments associated with the reuse of stormwater in residential areas. The most serious problems appear to be associated with the presence of potential pathogens in problematic numbers. Contact recreation in pathogen contaminated waters has been studied at many locations. The sources of the pathogens is typically assumed to be sanitary sewage effluent, or periodic industrial discharges from certain food preparation industries (especially meat packing and fish and shellfish processing). However, several studies have investigated pathogen problems associated with stormwater discharges. It has generally been assumed that the source of the pathogens in the stormwater are from inappropriate sanitary connections. However, as will be shown later, stormwater unaffected by these inappropriate sources still contains high counts of pathogens that are also found in surface runoff samples from many urban surfaces. Needless-to-say, sewage contamination of urban streams is an important issue that needs attention during an urban water resource investigation. Therefore, the following paragraphs present a brief overview of this contamination. Evidence of Sewage Contamination of Urban Streams The following case studies present summaries of various studies conducted throughout the U.S. that investigated contamination of urban streams that were only supposed to be receiving stormwater discharges. Many of the problematic discharges were from sanitary sewage. Obviously, inappropriate discharges must be identified and corrected as part of any effort to clean up urban streams. If these sources are assumed to be non-existent in an area and are therefore not considered in the stormwater management activities, incorrect and inefficient management decisions are likely, with disappointing improvements in the receiving waters. Lalor (1993), Pitt, et al. (1993), and Pitt and Lalor (1997) present a strategy to support the outfall screening activities required by the NPDES Stormwater Permit Program to identify and correct these inappropriate discharges to storm drainage systems. Nationwide A number of issues emerged from the individual projects of the U.S. EPA s Nationwide Urban Runoff Program (NURP) (EPA 1983). One of these issues involved illicit connections to storm drainage systems and was summarized as follows in the Final Report of the NURP executive summary: A number of the NURP projects identified what appeared to be illicit connections of sanitary discharges to stormwater sewer systems, resulting in high bacterial counts and dangers to public health. The costs and complications of locating and eliminating such connections may pose a substantial problem in urban areas, but the opportunities for dramatic improvement in the quality of urban stormwater discharges certainly exist where this can be accomplished. Although not emphasized in the NURP effort, other than to assure that the selected monitoring sites were free from sanitary sewage contamination, this BMP (Best Management Practice) is clearly a desirable one to pursue. The illicit discharges noted during NURP were especially surprising, because the monitored watersheds were carefully selected to minimize factors other than stormwater. Presumably, illicit discharge problems in typical watersheds would be much worse. Illicit entries into urban storm sewerage were identified by flow from storm sewer outfalls following substantial dry periods. Such flow could be the result of direct illicit connections as mentioned in the NURP final report, or could result from indirect connections (such as contributions from leaky sanitary sewerage through infiltration to the separate storm drainage). Many of these dry-weather flows are continuous and would therefore also occur during rain induced runoff periods. Pollutant contributions from the dry-weather flows in some storm drains have been shown to be high enough to significantly degrade water quality because of their substantial contributions to the annual mass pollutant loadings to receiving waters. Washtenaw County (Ann Arbor), MI 3

4 From 1984 to 1986, Washtenaw County, Michigan, dye-tested 160 businesses in an effort to locate direct illicit connections to the county stormwater sewerage (Murray 1985; Schmidt and Spencer 1986; Washtenaw County 1988). Of the businesses tested, 61 (38%) were found to have improper storm drain connections. The Huron River Pollution Abatement Program was the most thorough investigation of such improper connections. Beginning in 1987, 1067 businesses, homes and other buildings located in the Huron River watershed were dye tested. The following results were reported. Illicit connections were detected at 60% of the automobile related businesses inspected, including service stations, automobile dealerships, car washes, and auto body and repair shops. All plating shops inspected were found to have improper storm drain connections. Additionally, 67% of the manufacturers tested, 20% of the private service agencies, and 88% of the wholesale/retail establishments tested were found to have improper storm sewer connections. Of 319 homes dye tested, 19 were found to have direct sanitary connections to storm drains. The direct discharge of rug cleaning wastes into storm drains by carpet cleaners was also noted as a common problem. Several surveys, beginning as early as 1963, identified bacterial and chemical contamination of the Allen Creek storm drainage system. Studies in 1963, 1978 and 1979 found that discharges from the Allen Creek storm drain contained significant quantities of fecal coliform and fecal streptococci. The 1979 study also documented high pollutant loads of solids, nitrates and metals. A large number of inappropriate storm drain connections originating from businesses were found, especially within automobile related facilities. Chemical pollutants, such as detergents, oil, grease, radiator wastes and solvents were causing potential problems. The elimination of these storm drain connections prevented thousands of gallons of contaminated water from entering the Huron River from the Allen Creek storm drainage system annually. Eight sampling locations along the main stem and major lateral branches of the storm drainage system were established and monitored for 37 chemicals during rain events. From 1984 to 1986, 32 (86%) of these chemicals showed a decrease in concentrations while only 2 (5%) showed an increase. In spite of this improvement, chemical concentrations in the stormwater discharges at the Allen Creek outfall were still greater than those from the control station much of the time. Fort Worth, TX This program has been underway since June of 1985 (Falkenbury 1987). Investigations to date indicate few direct connections from industries to storm drains. Storm runoff, in addition to illegal dumping, accidental spills and direct discharges into the street or adjacent creeks seem to account for the majority of the contaminants entering the storm drainage system. Major problems stemmed from septic tanks, self-management of liquid wastes by industry and construction of municipal overflow bypasses from the sanitary sewer to the storm drains. The success of this program was judged by a decline in the number of undesirable features at the target outfalls. An average of 44 undesirable observations per month were made in 1986 (522 total), compared to an average of 21 undesirable observations per month in Inner Grays Harbor, WA In 1987, an inspection of the 90 urban stormwater outfalls draining into Inner Grays Harbor in Washington revealed 29 (32%) flowing during dry weather (Beyer, et al. 1979; Pelletier and Determan 1988). A total of 19 outfalls (21%) were described as suspect, based on visual observation and/or anomalous pollutant levels, as compared to those expected in typical urban stormwater runoff characterized by NURP. At least one storm drain system was later found to receive a residential sanitary sewage connection which has since been corrected. This drain exhibited no unusual visual characteristics, but was found to have atypical ph and total suspended solids levels. Notably, fecal coliform levels were within the typical range expected for stormwater. Sacramento, CA A Sacramento, California, investigation of urban discharges identified commercial as well as domestic discharges of oil and other automobile related fluids as a common problem based on visual observations (Montoya 1987). Montoya found that slightly less than half the water discharged from Sacramento's stormwater drainage system was not directly attributable to precipitation. Most of this water comes from unpermitted sources, including illicit and/or inappropriate entries to the storm drainage system. Bellevue, WA During the Bellevue, Washington Urban Runoff Project baseflows as well as stormwater from two residential urban basins were monitored (Pitt 1985; Pitt and Bissonnette 1983). The areas included in this study, Surrey Downs and 4

5 Lake Hills, are about 5 km apart and each covered an area of about 40 ha. Both were fully developed, with predominantly single family residences. No septic tanks were present in either area and the storm drainage systems were thoroughly mapped and investigated to ensure no non-stormwater discharges to storm drainage systems or obvious illegal discharges. The Bellevue, Washington, NURP project also summarized the reported incidents of intermittent discharges and dumpings of pollutants into the local storm drainage system. During a three year period of time, about 50 citizen contacts were made to the Bellevue Storm and Surface Water Utility District concerning water quality problems. About 25 percent of the complaints concerned oil being discharged into storm drain inlets. Another important category of complaints was for aesthetic problems, such as turbid or colored water in the creeks. Various industrial and commercial discharges into the storm drainage system were detected. Concrete wastes flushed from concrete trucks at urban job sites were a frequently occurring problem. Cleaning establishment discharges into creeks were also a common problem. Vehicle accidents also resulted in discharges of gasoline, diesel fuel, hydraulic fluids, and lawn care chemicals into the storm drain inlets. Boston, MA A field screening program was conducted to determine the relative levels of contamination at various locations in the Stony Brook drainage system (Metcalf and Eddy 1994). During eight days of dry-weather sampling, numerous inappropriate discharges of sanitary sewage into the drainage system were identified using the investigative procedures developed by Pitt, et al. (1993) and a modified flow chart approach. Minneapolis/St. Paul, MN Water Environment & Technology (1996a) reported that the fecal coliform counts decreased from about 500 counts/100 ml to about 150 counts/100 ml in the Mississippi River after the sewer separation program in the Minneapolis and St. Paul area of Minnesota. Combined sewers in 8,500 ha were separated during this 10-year, $332 million program. Toronto, Ontario The Toronto Area Watershed Management Strategy (TAWMS) study monitored and characterized both stormwater and baseflows (Pitt and McLean 1986 and GLA 1983). The project involved intensive monitoring in two test areas. The Emery catchment area, located near the City of North York, covered approximately 154 ha with predominantly medium industrial land uses (processing goods for final consumption). The Thistledown catchment, located in the City of Etobicoke, covered approximately 39 ha with residential and commercial land uses. During cold weather, the increases in dissolved solids were quite apparent in baseflows and snowmelt for both study catchments. This increase was probably caused by high chlorides from road salt applications. In contrast, bacteria populations were noticeably lower in all outfall discharges during cold weather. Nutrient and heavy metal concentrations at the outfalls remained fairly constant during cold and warm weather. Either warm- or cold-weather baseflows were responsible for most of the yields for many constituents from the industrial catchment. These constituents included runoff volume, phosphorus, total Kjeldahl nitrogen, chemical oxygen demand and chromium. Important constituents that had high yields in the baseflow from the residential/commercial catchment included total solids, dissolved solids, chlorides, and fecal coliform and Pseudomonas aeruginosa bacteria. Gartner Lee and Associates, Ltd. conducted an extensive survey of dry-weather flows in storm drainage systems in the Humber River watershed (Toronto) in an attempt to identify the most significant urban runoff pollutant sources. About 625 outfalls were sampled two times during dry-weather, with analyses conducted for many pollutants, including organics, solids, nutrients, metals, phenols, and bacteria. About 59% had dry-weather flows, and about 33% of the outfalls were discharging at rates greater than 1 L/sec. The dry-weather flows were found to contribute significant loadings of nutrients, phenols, and metals, compared to upstream conditions. About 10 to 14 percent of the outfalls were considered significant pollutant sources. Further investigations identified many industrial and sanitary sewage non-stormwater discharges into the storm drainage. An apartment building with the sanitary drains from eight units illegally connected to the storm drainage system was typical of the problems found. Other problem areas were found in industrial areas, including yard storage of animal hides and yard runoff from meat packing plants. 5

6 Ottawa, Ontario Visual inspection of stormwater pipes discharging to the Rideau River (Ontario) found leakage from sanitary sewer joints or broken pipes to be a major source of storm drain contamination (OME 1983). A study of the lower Rideau River in the Regional Municipality of Ottawa-Carleton was conducted to establish the causes of bacteriological water quality degradation in the urbanized reach of the river and to analyze the impacts of future urbanization. Earlier programs had identified and corrected many cross-connections between sanitary sewers and stormwater sewers. Bacteriological water quality improved, but swimming standards at beaches were still not obtained. Birmingham, AL During the development of the methods to investigate inappropriate discharges, a three-mile section of Village Creek in Birmingham, AL, was selected for field verification of the test methods (Pitt, et al. 1993, Pitt and Lalor 1997). The drainage area for this section of the creek contains about 4500 acres. Residential land use comprises approximately 88% of the total area, commercial land use approximately 8%, and industrial land use less than 1%. The majority of the drainage area is serviced by sanitary sewers, but some septic tanks are also used. A total of 65 stormwater outfalls were located. Outfall diameters ranged from 2 inches to 12 feet, excluding open ditches. All sites were visited at least 8 times during the field investigation period. Of these 65 outfalls, 48 (74%) were always dry, 6 (9%) had flow intermittently, and 11 (17%) were always flowing. Eighteen direct unpermitted discharges to the creek from nearby industries and commercial areas were also located; 10 (56%) were always dry, 6 (33%) had intermittent flow, and 2 (11%) were always flowing. The dry weather flows from two of the 65 outfalls were found to be mostly sanitary sewage, while the flows from another nine were predominately washwaters. The remaining outfalls with dry weather flows were mostly affected by natural waters (most likely groundwater infiltration) or leaking domestic water. Periodic stream surveys of tributaries of the Cahaba River in the Birmingham area (mostly the Little Cahaba River, upstream of Lake Purdy) during summer months have found that the small river contained about 1/3 treated sewage from upstream poorly operated municipal treatment facilities (since corrected), septage from failing septic tanks, and SSO discharges. During a current EPA sponsored project investigating SSO discharges being conducted by Lalor at UAB, sewage, through SSOs and poorly operating septic tanks, were found to make up about 25% of the dry weather flows in the small, completely urbanized stream in Homewood, AL, being studied. However, sewage contributions in the much larger, and much less urbanized 5-mile Creek are very small (on a percentage standpoint), although SSOs exist in the urbanized area. These streams are still being evaluated, including future human health risk assessments associated with these discharges. Summary of Inappropriate Sanitary Sewage Discharges into Urban Streams Urban stormwater runoff includes waters from many other sources which find their way into storm drainage systems, besides from precipitation. There are cases where pollutant levels in storm drainage are much higher than they would otherwise be because of excessive amounts of contaminants that are introduced into the storm drainage system by various non-stormwater discharges. Additionally, baseflows (during dry weather) are also common in storm drainage systems. Dry-weather flows and wet-weather flows have been monitored during numerous urban runoff studies. These studies have found that discharges observed at outfalls during dry weather were significantly different from wet-weather discharges and may account for the majority of the annual discharges for some pollutants of concern from the storm drainage system. In many cases, sanitary sewage was an important component (although not necessarily the only component) of the dry weather discharges from the storm drainage systems. From a human health perspective (associated with pathogens), it may not require much raw or poorly treated sewage to cause a receiving water problem. However, at low discharge rates, the DO receiving water levels may be minimally affected. The effects these discharges have on the receiving waters is therefore highly dependent on many site specific factors, including frequency and quantity of sewage discharges and the creek flows. In many urban areas, the receiving waters are small creeks in completely developed watersheds. These creeks are the most at risk from these discharges as dry base flows may be predominately dry weather flows from the drainage systems. In Tokyo (Fujita 1998), for example, numerous instances were found where correcting inappropriate sanitary sewage discharges resulted in the urban streams losing 6

7 all of their flow. In cities that are adjacent to large receiving waters, these discharges likely have little impact (such as DO impacts from Nashville CSO discharges on the Cumberland River, Cardozo, et al. 1994). The presence of pathogens from raw, or poorly treated sewage, in urban streams, however, obviously presents a potentially serious public health threat. Even if the receiving waters are not designated as water contact recreation, children are often seen playing in small streams in urban areas. Epidemiological Studies and Human Exposures to Stormwater (after Craun, et al. 1996) Epidemiology can be defined as the study of the occurrence and causes of disease in human populations and the application of this knowledge to the prevention and control of health problems. The general population often views epidemiology and associated risk assessments with skepticism when risks associated with seemingly everyday activities are quantified, especially when associated with periodic food scares that are typically exaggerated or misinterpreted in the press. Technical experts also may feel uncomfortable with the results of epidemiological studies because of the typically very low numbers of affected people in a study population. However, much of the information that is used in developing environmental regulations protecting human health originates with epidemiological studies and a more through understanding of the science of epidemiology would dispel much of the confusion associated with these studies. Epidemiology has routinely been used to assess risks associated with contaminants in drinking waters. Epidemiology has also recently been used to investigate human health risks associated with swimming in waters contaminated by stormwater. However, Craun, et al. (1996) state that the results of environmental epidemiology studies (the assessment of human health effects associated with environmental contaminants, where indicators of disease are mostly studied instead of the disease itself) have provoked controversy. Their excellent review article on epidemiology applied to water and public health discusses many of these problems and offers suggestions to enable better interpretation of existing studies and better design of future studies. The following paragraphs are summarized from their article. The definition of terms is important. For example, epidemiologists use several measures to describe disease frequency. Incidence is the rate at which new cases of disease occur, whereas prevalence measures both new and existing cases in the total population. Therefore, prevalence is the proportion of people who have a specific condition at any specific time and is typically measured as a percentage of the total population. Incidence considers the duration of exposure, and the incidence rate may be expressed as the number of cases observed per person-years of exposure, for example. The attack rate is a measure of the cumulative incidence during an outbreak of the disease, and is usually expressed in terms of numbers of cases of disease per population unit (such as per 10,000 people in the population). Secondary outbreaks can also occur for communicable disease and the secondary attack rate refers to the cases of disease attributed to exposure to people having the disease during the primary attack. The secondary rate is usually expressed in terms of susceptible contacts. Geographic-specific (such as part of town receiving water from a specific source) and vehicle-specific (such as waterborne specific disease) attack rates help to determine the source of the disease. Attack rates can also be examined in terms of water consumption by separating the attack rate into different categories associated with different amounts of water consumed, for example. Mortality rate and case fatality rate are also measures of disease frequency. The mortality rate indicates the number of deaths from a certain disease, and or time period, per the total population. The case fatality rate is the proportion of diagnosed individuals having the disease who die of the disease. The crude rates should be standardized to account for differences in demographic characteristics of the population, especially age. Association is a measure of the dependence between exposure to a contaminant and the onset of disease, but does not necessarily indicate a cause and effect relationship between the variables. Both experimental (clinical or population) and observational (descriptive or analytical) epidemiologic studies are used to determine associations. In clinical experimental studies, active intervention may be used to expose the subjects to specific doses of an infective agent to determine the infective dose of a pathogen, for example. In population experimental studies, the population may be randomly grouped according to different levels of drinking water treatment, and the households would then be extensively examined to determine any differences in disease outbreaks. In descriptive observational studies, information is available about the occurrence of disease and about exposure to specific compounds, exposure periods, and different demographic information. Analytical observational studies test specific hypotheses to evaluate associations between exposure and disease and to confirm the mode of transmission. Ecological studies (or 7

8 correlation or aggregate studies) examine associations between routinely gathered health and demographic statistics and available environmental measures (such as drinking water constituent concentrations). These studies are typically controversial because the statistical demographic information pertains to groups (lumped information which makes it difficult to identify confounding factors or to normalize) and not to individuals within the groups. Difficulties also relate to incomplete information concerning potential causative agents. Therefore, analytical observational studies (where individuals are studied and more detailed information concerning the potential causative agents can be obtained) should be used to follow up hypotheses developed in ecological studies. The experimental design of epidemiological studies is very critical. The study must be of sufficient size and have adequate statistical power to detect the hypothesized association. Randomness is also very critical in epidemiological studies to control systematic errors. In most cases, epidemiological studies compare disease rates between a test and a control population. Positive associations (where there is a statistically significant difference between the rates of the two groups) can be caused by random errors. This likelihood can be estimated by calculating the confidence interval of the statistical significance of the association. However, statistical significance (even at a very high level) does not imply a cause and effect relationship between the hypothesized factor and disease. Statistical power can be used to identify the minimum risk that a study is capable of detecting. An environmental epidemiological study should not be conducted unless the exposure assessment is expected to be reasonably appropriate and accurate. Adequate and complete data to make the exposure assessment must be assured before the study is conducted. Interpreting associations is based on examining the rate differences (RD), which is the absolute differences in the two rates (incidence rate of disease for the test, or exposed, group minus the incidence rate of disease for the control, or unexposed, group), or the rate ratio (RR), which is the ratio of the rates from the two groups. The odds ratio (OR) is the ratio of the odds of disease of the test group to the odds of disease of the control group, and is interpreted similarly to the rate ratio. If the RR or OR is close to 1.0, there is no association or increased risk between the two groups. If the ratio is 1.8, there is an 80 percent increased risk of disease for the exposed individuals, compared to the unexposed group. The confidence interval of the ratio is used to identify significance of the association. A 95 percent confidence interval of 1.6 to 2.0 signifies a statistically significant estimate because the range does not include 1.0. The relatively narrow range also implies a precise estimate of the association. In contrast, a 95 percent confidence interval of 0.8 to 14.5 does not signify a significant difference because the range includes the value of 1.0. In addition, the wide range also implies an imprecise estimate of the association. Craun, et al. (1996) presents Table 1 (from Monson 1980) indicating different rate ratios and strengths of associations. Weak associations (ratios of <1.5) are difficult to interpret. Very large range ratios are unlikely to be completely explained by unidentified or uncontrolled confounding characteristics. However, the magnitude of the rate ratio has no bearing on the likelihood that the association is attributed to bias, but causal association cannot be ruled out simply because of a weak association. In many environmental epidemiological studies, the rate ratio is frequently smaller than 1.5, causing speculation that the association may actually be caused by bias. High quality exposure and study design are important for interpreting risks of this magnitude. Table 1. Rate Ratios and Strengths of Associations for Epidemiological Studies (Monson 1980) Rate Ratio, or Strength of Association Odd Ratio 1.0 None >1.0 to <1.5 Weak 1.5 to 3.0 Moderate 3.1 to 10.0 Strong >10.0 Infinite With the low rate ratios frequently encountered in environmental epidemiological studies, cautious interpretations are necessary. Craun, et al. (1996) present the following criteria that are used to assess associations and causality: 8

9 Exposure must occur before the onset of disease (temporal association) A sufficient number of participants are needed to prevent random error, and the study is well conducted (study precision and validity) The range ratio (or odds ratio) should be large enough to minimize spurious associations (strength of association) Repeated observations are needed under different conditions to support causality (consistency) The absence of specificity does not rule out causality, but a commonly accepted effect associated with a specific exposure certainly reinforces causality (specificity) An association supported by scientific evidence supports causality (biological plausibility) Higher risks should be associated with higher exposures (dose-response relationship) The removal of a potential causative agent should reduce the risk of disease (reversibility) Therefore, an effective and convincing interpretation can be supported if many of these above factors are successfully addressed by an environmental epidemiological study. Water Contact Recreation and Urban Stormwater There have been a few epidemiology studies recently published describing the increased health risks associated with contaminated dry weather flows affecting public swimming beaches. The following discussion presents an overview of the development of water quality criteria for water contact recreation, plus the results of several epidemiological studies that have specifically examined human health problems associated with swimming in contaminated water, including water affected by stormwater. In most cases, the levels of indicator organisms and pathogens causing increased illness were well within the range found in urban streams. Development of Bathing Beach Bacteriological Criteria and Associated Epidemiological Studies Human health standards for body contact recreation (and for fish and water consumption) are based on indicator organism monitoring. Monitoring for the actual pathogens, with few exceptions, requires an extended laboratory effort, is very costly and not very accurate. Therefore, the use of indicator organisms has become established. Dufour (1984a) presents an excellent overview of the history of indicator bacterial standards and water contact recreation, summarized here. Total coliforms were initially used as indicators for monitoring outdoor bathing waters, based on a classification scheme presented by W.J. Scott in Total coliform bacteria refers to a number of bacteria including Escherichia, Klebsiella, Citrobacter, and Enterobacter (DHS 1997). They are able to grow at 35 o C and ferment lactose. They are all gram negative asporogenous rods and have been associated with feces of warm blooded animals. They are also present in soil. Scott had proposed four classes of water, with total coliform upper limits of 50, 500, 1,000, and >1,000 MPN/100 ml for each class. He had developed this classification based on an extensive survey of the Connecticut shoreline where he found that about 93% of the samples contained less than 1,000 total coliforms per 100 ml. A sanitary survey classification also showed that only about 7% of the shoreline was designated as poor. He therefore concluded that total coliform counts of <1,000 MPN/100 ml probably indicated acceptable waters for swimming. This standard was based on the principle of attainment, where very little control or intervention would be required to meet this standard. In 1943, the state of California independently adopted an arbitrary total coliform standard of 10 MPN/1 ml (which is the same as 1,000 MPN/100 ml) for swimming areas. This California standard was not based on any evidence, but it was assumed to relate well with the drinking water standard at the time. Dufour points out that a third method used to develop a standard for bathing water quality used an analytical approach adopted by H.W. Streeter in He used a ratio between Salmonella and total coliforms, the number of bathers exposed, the approximate volume of water ingested by bathers daily, and the average total coliform density. Streeter concluded that water containing <1,000 MPN total coliforms/100 ml would pose no great Salmonella typhosa health hazard. Dufour points out that it is interesting that all three approaches in developing a swimming water criterion resulted in the same numeric limit. One of the earliest bathing beach studies to measure actual human health risks associated with swimming in contaminated water was directed by Stevenson (1953), of the U.S. Public Health Service s Environmental Health Center, in Cincinnati, Ohio, and was conducted in the late 1940s. They studied swimming at Lake Michigan at 9

10 Chicago (91 and 190 MPN/100 ml median total coliform densities), the Ohio River at Dayton, KY (2,700 MPN/100 ml), at Long Island Sound at New Rochelle and at Mamaroneck, NY (610 and 253 MPN/100 ml). They also studied a swimming pool in Dayton, KY. Two bathing areas were studied in each area, one with historically poorer water quality than the other. Individual home visits were made to participating families in each area to explain the research program and to review the calendar record form. Follow up visits were made to each participating household to insure completion of the forms. Total coliform densities were monitored at each bathing area during the study. More than 20,000 persons participate in the study in the three areas. Almost a million person-days of useable records were obtained. The percentage of the total person-days when swimming occurred ranged from about 5 to 10 percent. The number of illnesses of all types recorded per 1,000 person-days varied from 5.3 to 8.8. They found an appreciably higher illness incidence rate for the swimming group, compared to the nonswimming group, regardless of the bathing water quality (based on total coliform densities). A significant increase in gastrointestinal illness was observed among the swimmers who used one of the Chicago beaches on three days when the average coliform count was 2,300 MPN/100 ml. The second instance of positive correlation was observed in the Ohio River study where swimmers exposed to the median total coliform density of 2,700 MPN/100 ml had a significant increase in gastrointestinal illness, although the illness rate was relatively low. They suggested that the strictest bacterial quality requirements that existed then (as indicated above, based on Scott s 1934 work) might be relaxed without significant detrimental effect on the health of bathers. It is interesting to note that in 1959, the Committee on Bathing Beach Contamination of the Public Health Laboratory Service of the UK concluded that bathing in sewage-polluted seawater carries only a negligible risk to health, even on beaches that are aesthetically very unsatisfactory (Cheung, et al and Alexander, et al. 1992). Dufour (1984a) pointed out that total coliforms were an integral element in establishing fecal coliform limits as an indicator for protecting swimming uses. Fecal coliform bacteria are a subgroup of the total coliform group. They grow at 44.5 o C and also ferment lactose. They are restricted to the feces of warm blooded animals and can be used to separate bacteria of soil and animal origin (DHS 1997). They do survive for variable periods of time in fecal contaminated soil and water, however. As a result of the Stevenson (1953) study, reported above, a geometric mean fecal coliform level of 200 MPN per 100 ml was recommended by the National Technical Advisory Committee (NTAC) of the Federal Water Pollution Control Administration in 1968 and was adopted by the U.S. Environmental Protection Agency in 1976 as a criterion for direct water contact recreation (Cabelli, et al. 1979). This criterion was adopted by almost all states by It was felt that fecal coliforms was more specific to sewage contamination and had less seasonal variation that total coliforms. Since fecal coliform exposures at swimming beaches had never been linked to disease, the NTAC reviewed the USPHS studies, as published by Stevenson (1953). The 2,300 MPN/100 ml total coliform count association with gastrointestinal disease was used in conjunction with a measured ratio of fecal coliform to total coliform counts (18%) obtained at the Ohio River site studied earlier. It was therefore assumed that a health effect could be detected when the fecal coliform count was 400 MPN/100 ml (18% of 2,300 = 414). Dufour (1984a) pointed out that a detectable health effect was undesirable and that the NTAC therefore recommended a limit of 200 MPN/100 ml for fecal coliforms. Dufour (1984a) points out that, although likely coincidental, the 1968 proposed limit for fecal coliforms (200 MPN/100 ml) was very close to being theoretically equivalent to the total coliform limit of 1,000 MPN/100 ml that was being replaced (200/0.18 = 1100). Dufour (1984a) lists the ideal characteristics of bacterial indicators of fecal contamination, as presented by various authors. The authors were in agreement concerning many of the criteria (correlation to pathogens, unable to grow in aquatic environments, more resistant to disinfection than pathogens, and easy to isolate and enumerate), but two important aspects were seldom mentioned, namely that the indicator should have a direct relationship to fecal contamination, and that the indicator density should correlate with health hazards. Many of the follow-up studies conducted since the mid 1970s examined these additional criteria. E. coli, a member of the fecal coliform group, has been used as a better indicator of fresh fecal contamination. Table 2 indicates the species and subspecies of the Streptococcus and Enterococcus groups of bacteria that are used as indicators of fecal contamination (DHS 1997). Fecal streptococci bacteria are indicators of fecal contamination. The enterococcus group is a subgroup that is considered a better indication of human fecal contamination. S. bovis and S. equinus are considered related to feces 10

11 from non-human warm blooded animals (such as from meat processing facilities, dairy wastes, and feedlot and other agricultural runoff), indicating that enterococcus may be a better indication of human feces contamination. However, S. facealis subsp. liquifaciens is also associated with vegetation, insects, and some soils (DHS 1997). Table 2. Streptococcus Species used as Indicators of Fecal Contamination Indicator organism Enterococcus group Group D antigen Streptococcus faecalis X X S. facealis subsp. liquifaciens X X S. faecalis subsp. zymogenes X X S. faecium X X S. bovis X S. equinus X Group Q antigen S. avium X Source: DHS (1997) Streptococcus group The Cabelli, et al. (1979) study was undertaken to address many remaining questions pertaining to bathing in contaminated waters. Their study examined conditions in New York (at a Coney Island beach, designated as barely acceptable, and at a Rockaway beach, designated as relatively unpolluted). About 8,000 people participated in the study, approximately evenly divided between swimmers and nonswimmers at the two beaches. Total and fecal coliforms, Escherichia, Klebsiella, Citrobacter-Enterobacter, Enterococci, Pseudomonas aeruginosa, and Clostridium perfringens were evaluated in water samples obtained from the beaches during the epidemiological study. The most striking findings were the increases in the rates of vomiting, diarrhea, and stomachache among swimmers relative to nonswimmers at the barely acceptable beach, but not at the relatively unpolluted beach. Ear, eye, nose, and skin symptoms, as well as fever, were higher among swimmers compared to nonswimmers at both beaches. They concluded that measurable health effects do occur at swimming beaches that meet the existing health standards. Children, Hispanic Americans, and low-middle socioeconomic groups were identified as the most susceptible portions of the population. Cabelli, et al. (1982) presented data from the complete EPA sponsored swimming beach study, conducted in New York, New Orleans, and Boston. The study was conducted to address issues from prior studies conducted in the 1950s (including Stevenson s 1953 study noted above) that were apparently contradictory. They observed a direct, linear relationship between highly credible gastrointestinal illness and enterococci. The frequency of gastrointestinal symptoms also had a high degree of association with distance from known sources of municipal wastewater. Table 3 shows correlation coefficients for total gastrointestinal (GI) and highly credible gastrointestinal (HCGI) symptoms and mean indicator densities found at the New York beaches from 1970 to The best correlation coefficients were found for enterococci. In contrast, the correlation coefficients for fecal coliforms (the basis for most federal and state guidelines) were poor. Very low levels of enterococcus and Escherichia coli in the water (about 10 MPN/100 ml) were associated with appreciable attack rates (about 10/10,000 persons). Table 3. Correlation Coefficients between Gastrointestinal Symptoms and Bacterial Densities at New York City Beaches (Cabelli, et al. 1982) Indicator HCGI correlation coefficient GI correlation coefficient Number of observations Enterococci Escherichia coli Klebsiella Enterobacter-Citrobacter Total coliforms

12 Clostridium perfringens Pseudomonas aeruginosa Fecal coliforms Aeromonas hydriphila Vibrio parahemoylticus Figure 1 shows regressions of swimming associated gastrointestinal symptom rates (swimmer rates minus nonswimmer rates) against the mean enterococcus and E. coli densities of the water samples. The results clearly show that the risk of gastrointestinal symptoms associated with swimming in marine waters contaminated with municipal wastewater is related to the quality of the water, as indicated by the enterococcus density of the water. They also felt there was a strong case for causality between enterococci and gastrointestinal symptoms, based on the good association, the consistency at the different locations over different years, the reasonable nature of the relationship between enteric disease and fecal contamination, and the coherent association based on observations of waterborne disease transmission during prior outbreaks. They concluded that swimming in even marginally polluted marine bathing water is a significant route of transmission for observed gastrointestinal illness. They felt that the gastrointestinal illness was likely associated with the Norwalk-like virus that had been confirmed in 2,000 cases at a shellfish associated outbreak in Australia and at several outbreaks associated with contaminated drinking water. Fleisher (1991) reevaluated this marine swimming beach data and concluded that the limitation for enterococci promulgated by the EPA in 1986, based on the Cabelli, et al. (1982) study, (35 per 100 ml, geometric mean for 5 equally spaced samples over a 30-day period, for both fresh and saline water) was too severe, due to minor adjustments of the observed data. He was also especially concerned with the use of a single criterion based on pooled data, while the data from the individual sites indicated very different probabilities of gastroenteritis among swimmers at Boston compared to New York and Lake Pontchartrain (which were similar). He also reported that previous studies found bacteria indicator, and possibly pathogen, survival to be inversely correlated with salinity. He therefore concluded that any relation between enterococci and disease causing pathogens may be site specific, possibly related to water salinity. This EPA enterococci criterion for swimming waters was based on an acceptable rate of gastroenteritis of 19 cases per 1,000 swimmers, the same rate upon which the fecal coliform criterion (200 MPN/100 ml) was based. It is interesting to note that Fleisher later participated in additional epidemiological studies in the UK and concluded that 33 fecal streptococci (essentially enterococci)/100 ml was the threshold of increased risk for gastrointestinal illness for swimmers (Kay, et al. 1994). Dufour (1984a) also reviewed a series of studies conducted at freshwater swimming beaches from 1979 to 1982, at Tulsa, OK, and at Erie, PA. Only enterococci, E. coli, and fecal coliforms were monitored, based on the results of the earlier studies. Table 4 shows the correlation coefficients for these three bacterial parameters and gastrointestinal disease. Table 4. Correlation Coefficients for Bacterial Parameters and Gastrointestinal Disease (Fresh Water Swimming Beaches) Highly Credible Gastrointestinal Illness Total Gastrointestinal Illness Enterococci E. coli Fecal coliforms Number of Study Units These results are quite different than the results from the marine studies, in that both enterococci and E. coli had high correlation coefficients between the bacterial levels and the incidence of gastrointestinal illness. However, the result was the same for fecal coliforms, in that there was no association between fecal coliform levels and gastrointestinal illness. Dufour (1984b) concluded that enterococci would be the indicator of choice for gastrointestinal illness, based on scientific dependability. E. coli could also be used, if only fresh waters were being evaluated. Fecal coliforms would be a poor choice for monitoring the safety of bathing waters. However, he 12